International Journal of Advances in Engineering and Management (IJAEM)

Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

Optimal power flow using Modified

Driving-Training Based Optimization
algorithm
O. M. Ranarison1, E. Randriamora2, H. Andriatsihoarana3
1
Ph.D Student in Electrical Engineering, Ecole Doctorale en Sciences et Techniques de l’Ingénierie et de
l’Innovation (ED-STII), University of Antananarivo, Antananarivo, Madagascar.
2
Thesis Co-Director, ED-STII, University of Antananarivo, Antananarivo, Madagascar.
3
Thesis Director, ED-STII, University of Antananarivo, Antananarivo, Madagascar.
Corresponding Author: Olivier Mickaël Ranarison,

--------------------------------------------------------------------------------------------------------------------------- ------------
Date of Submission: 25-02-2025 Date of Acceptance: 05-03-2025
--------------------------------------------------------------------------------------------------------------------------- ------------

ABSTRACT: The control and study of electric I. INTRODUCTION

power system require solution of Optimal Power OPF is firstly introduced by Carpentier in
Flow (OPF) problem which is considered one of the 1962 [1]. The goal of OPF is to optimize a selected
most difficult optimization problems. The objective objective functions such as generation fuel cost,
of OPF is finding the most secure operating point or voltage profile improvement, voltage stability
the best control variables while considering equality enhancement, minimization of active power
and inequality constraints of the system, and transmission losses, reactive power losses
optimizing certain objective functions. In this paper, minimization, by adjusting the power system control
five objective functions are considered: variables which are the generator real powers, the
minimization of total generation fuel cost, voltage generator bus voltages, the transformer tap settings,
profile improvement, voltage stability enhancement, and the reactive power of switchable VAR sources.
minimization of active power transmission losses With considering power flow equations (equality
and reactive power losses minimization. Modified constraints) and the limits on control variables
Driving-Training Based Optimization (MDTBO) (inequality constraints), OPF becomes a large-scale
algorithm is used to solve the OPF problem. highly constrained nonlinear and nonconvex
Driving-Training-Based Optimization (DTBO) is a optimization problem. In the literature, Ebeed et al
human-based metaheuristic algorithm based on the (2018) classify all methods of OPF solution in two
simulation of driving training process. With main groups: conventional and recent optimization
MDTBO, a new method for choosing the number of methods [2]. Conventional methods are
driving learners and instructors is introduced. To deterministic and classic heuristic optimization
evaluate the proposed method, the standard IEEE techniques, such as linear, nonlinear and quadratic
30-bus network is used and results is compared to programming methods, Newton’s method and
another metaheuristic algorithm such as Teaching- interior point method, which are fully reviewed by
Learning Based Optimization (TLBO) and Particle Momoh et al. (1999) [3][4]. Recent optimization
Swarm Optimization (PSO) algorithm. The results methods are metaheuristics and can be classified in
show that the proposed approach is competitive with swarm and bio-inspired, human inspired,
other algorithms. evolutionary inspired, physics inspired and hybrid
optimization techniques, Artificial neural networks
KEYWORDS: Optimal Power Flow (OPF), (ANN) and fuzzy logic approaches. According to
generation cost, voltage stability, voltage profile, the state of the art about modern techniques by
active power transmission losses, reactive power Emmanuel et al (2021) [5] and Risi et al. (2022)[6],
losses, Teaching-Learning Based Optimization Attia et al. (2012) proposed an adapted genetic
(TLBO), Particle Swarm Optimization (PSO), algorithm with adjusting population size, applied in
Modified Driving-Training Based Optimization IEEE-30 bus system with fuel cost and voltage
(MDTBO). profile improvement as objective functions [7].The
work of Adaryani et al. (2013) studies the

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 846
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

application of Artificial bee colony algorithm for optimizing fuel cost, quadratic cost with valve-point
solving multi-objective OPF problem (fuel cost, effect, piecewise quadratic cost, emission, voltage
quadratic cost with valve-point effect, piecewise profile and stability, active power transmission
quadratic cost, emission, voltage stability, active losses, tested on IEEE 30-bus, 57-bus, 118-bus [18];
power transmission losses, voltage profile) on IEEE Modified coyote optimization algorithm by Li et al.
9-bus,30-bus and 57-bus systems [8]. Kahourzade et (2019) for minimization of fuel cost and active
al. (2015) proposed a comparative study of multi- power transmission losses on IEEE 30-bus [19];
objective OPF (fuel cost, emission minimization, Gorilla troops optimization technique by Shaheen et
voltage stability, active power transmission losses) al. (2022) for fuel cost, emission, and active power
based on particle swarm (PSO), evolutionary transmission losses minimization, on IEEE 30-bus
programming, and genetic algorithm, tested on and Egypt Power System [20]. Improved stochastic
IEEE-30 bus system [9]. Le Anh et al. (2015) study fractal search algorithm is a physic inspired
Cuckoo Optimization Algorithm for OPF with fuel algorithm proposed by Nguyen et al. (2020) for fuel
cost, quadratic cost with valve-point effect, and cost, emission, power losses minimizations, voltage
piecewise quadratic cost as objective functions, profile and stability on IEEE 30-bus, IEEE 57-bus,
tested on IEEE 30-bus, 57-bus and 118-bus system IEEE 118-bus systems [21]. Yuan et al. (2017) work
[10]. El-Fergany et al. (2015) are interested on grey on improved strength Pareto evolutionary algorithm
wolf optimizer and differential evolution algorithms for generation cost and emission minimization on
for minimizing generation fuel cost with valve-point IEEE 30-bus and 57-bus [22]. Fuzzy logic approach
effect, reactive power losses and active power for minimizing generation of fuel cost on IEEE 14-
transmission losses, on IEEE 30-bus and IEEE 118- bus is introduced by Ramesh et al. (1997) [23] and
bus [11]. The literature provides another human Artificial neural network for voltage stability and
inspired algorithms applied on OPF problem, such minimization of generation fuel cost on IEEE 30-
as Teaching-Learning Based Optimization (TLBO) bus is developed by Venkatesh (2003) [24]. Yuhao
introduced by Bouchekara et al. (2014) where IEEE et al. (2020) proposed Deep reinforcement learning
30-bus and 118-bus is tested with quadratic fuel for minimizing fuel cost on IEEE 14-bus and Illinois
cost, piecewise quadratic cost, voltage stability, 200-bus system [25]. Singh et al. (2022) study
voltage profile and active power transmission losses Sensitivity-Informed Deep Neural Networks for
as objective functions [12]. A modified version of minimizing fuel cost on IEEE 39-bus, IEEE 118-bus
TLBO for OPF is proposed by Shabanpour- and Illinois 200-bus [26]. In 2023, Xiang Pan
Haghighi et al. (2014) where objective functions are proposed a feasibility-optimized deep neural
quadratic fuel cost with valve-point effect and network for minimizing fuel cost on IEEE 30-bus,
emission minimization, and test systems are IEEE 118-bus, 300-bus and synthetic 2000-bus system
30-bus and 57-bus [13]. An improved teaching- [27]. After studying state of the art, no-one has
learning based optimization algorithm using Levy studied MDTBO approach for solving OPF
mutation strategy for non-smooth optimal power problem.
flow is also proposed by Ghasemi et al. (2015), In this article, the main objective is to
tested on IEEE 30-bus and 57-bus systems with modify original DTBO by introducing a new
voltage stability, emission minimization, generation equation for setting number of driving learners and
cost, quadratic cost with valve-point effect and instructors, and for optimizing five objective
piecewise quadratic cost, as objective functions functions of OPF: minimization of total generation
[14]. Duman (2016) develop Symbiotic organisms fuel cost (case 1), voltage profile improvement (case
search algorithm for OPF based on valve-point 2), voltage stability enhancement (case 3),
effect and prohibited zones, tested on IEEE 30-bus minimization of active power transmission losses
[15]. Adaptive multiple teams perturbation-guiding (case 4) and reactive power losses minimization
Jaya is proposed by Warid (2020) for generation (case 5).
cost minimization, voltage stability enhancement
and minimization of active power transmission II. MATERIALS AND METHODS
losses on IEEE 30-bus and 118-bus [16]. Warid Used material for this article is IEEE 30-
(2022) is also interested on Novel chaotic Rao-2 for bus system which is composed of six power
OPF solution, with four objective functions (fuel generators at buses 1, 2, 5, 8, 11 and 13; four
cost, voltage profile and stability, active power transformers with off-nominal tap ratio at lines 6-9,
transmission losses) tested on IEEE 30-bus and 118- 6-10, 4-12, and 28-27; and nine shunt VAR
bus [17]. New bio-inspired algorithms for OPF are compensation buses at buses 10, 12, 15, 17, 20, 21,
also proposed since 2017 such as Moth swarm 23, 24 and 29 [11]. Buses are coded as 1 for slack
algorithm introduced by Mohamed et al. (2017) for bus, 2 for PV bus and 3 for PQ bus (Fig. 1).

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 847
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

Bus data, generator data, line data and cost Data is manipulated in per unit with base
coefficients are giving in [28] [29]. The minimum equal to 100MVA. Precision’s value is 0.001.
and maximum limits for the control variables and Newton-Raphson method is used to determine load
for the line power transmissions are respectively flow parameters in the IEEE 30-bus system.
tabulated in (Table 1) and (Table 2).

Fig.1: IEEE 30-bus system

Table 1: Limits for the control variables

Variables Min Max
P1 50 200
P2 20 80
P5 15 50
P8 10 35
P11 10 30
𝑃13 12 40
𝑉1 0.95 1.1
𝑉2 0.95 1.1
𝑉5 0.95 1.1
𝑉8 0.95 1.1
𝑉11 0.95 1.1
𝑉13 0.95 1.1
𝑇11 0.9 1.1
𝑇12 0.9 1.1
𝑇15 0.9 1.1
𝑇36 0.9 1.1
𝑄10 0.0 5.0
𝑄12 0.0 5.0
𝑄15 0.0 5.0

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 848
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

𝑄17 0.0 5.0

𝑄20 0.0 5.0
𝑄21 0.0 5.0
𝑄23 0.0 5.0
𝑄24 0.0 5.0
𝑄29 0.0 5.0

Table 2: Limit for the line power transmissions

Line No From bus To bus S (p.u)
1 1 2 1.300
2 1 3 1.300
3 2 4 0.650
4 3 4 1.300
5 2 5 1.300
6 2 6 0.650
7 4 6 0.900
8 5 7 0.700
9 6 7 1.300
10 6 8 0.320
11 6 9 0.650
12 6 10 0.320
13 9 11 0.650
14 9 10 0.650
15 4 12 0.650
16 12 13 0.650
17 12 14 0.320
18 12 15 0.320
19 12 16 0.320
20 14 15 0.160
21 16 17 0.160
22 15 18 0.160
23 18 19 0.160
24 19 20 0.320
25 10 20 0.320
26 10 17 0.320
27 10 21 0.320
28 10 22 0.320
29 21 22 0.320
30 15 23 0.160
31 22 24 0.160
32 23 24 0.160
33 24 25 0.160
34 25 26 0.160
35 25 27 0.160
36 28 27 0.650
37 27 29 0.650
38 27 30 0.160
39 29 30 0.160
40 8 28 0.320
41 6 28 0.320

2.1. Optimal Power Flow formulation dependent variables, and 𝑢, vector of control
OPF problem can be modelled by minimizing an variables.
objective function 𝐹 composed of 𝑥, vector of 𝑀𝑖𝑛𝐹(𝑥, 𝑢) (1)
DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 849
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

with 𝑉𝑖 and 𝑉𝑗 are the voltage of 𝑖th generator bus and the
𝑔𝑗 (𝑥, 𝑢) = 0 ; 𝑗 = 1,2, … , 𝑚 (2) voltage of 𝑗th load bus, respectively.𝐹𝑖𝑗 can be
ℎ𝑗 (𝑥, 𝑢) ≤ 0 ; 𝑗 = 1,2, … , 𝑝 (3) expressed with submatrices of𝑌𝑏𝑢𝑠 matrix:
𝑔𝑗 and ℎ𝑗 represent equality and inequality 𝐹𝑖𝑗 = [𝐹𝐿𝐺 ] − [𝑌𝐿𝐿 ]−1 [𝑌𝐿𝐺 ] (10)
constraints; 𝑚and𝑝 represent the number of each
constraint. Case 4: Minimization of active power
𝑥 can be expressed as: transmission losses
𝑥 (4 This can be formulated as follows:
𝑁𝑇𝐿 (11)
= [𝑃𝐺1 , 𝑉𝐿1 … 𝑉𝐿,𝑁𝑃𝑄 , 𝑄𝐺1 … 𝑄𝐺,𝑁𝐺 , 𝑆𝑇𝐿,1 … 𝑆𝑇𝐿,𝑁𝑇𝐿 ])
where: 𝑃𝑙𝑜𝑠𝑠 = ∑ 𝐺𝑖𝑗 (𝑉𝑖2 + 𝑉𝑗2 − 2𝑉𝑖 𝑉𝑗 𝑐𝑜𝑠𝛿𝑖𝑗 )
𝑃𝐺1 , active power output at slack bus; 𝑉𝐿 , voltage 𝑖=1
magnitude at PQ buses ; 𝑄𝐺 , reactive power output 𝐺𝑖𝑗 conductance of a line, NTL number of
of all generator units ; 𝑆𝑇𝐿 , transmission line transmission lines, and 𝛿𝑖𝑗 phase difference of
loading; 𝑁𝑃𝑄 , number of load buses ; 𝑁𝐺, number voltages.
of generator units;𝑁𝑇𝐿, number of transmission
lines. Case 5: Reactive power losses minimization
𝑢 can be expressed as: This can be expressed by:
𝑁𝑇𝐿 (12)
𝑢 (5
= [𝑃𝐺,2 … 𝑃𝐺,𝑁𝐺 , 𝑉𝐺,1 … 𝑉𝐺,𝑁𝐺 , 𝑄𝐶,1 … 𝑄𝐶,𝑁𝐶 , 𝑇1 … 𝑇𝑁𝑇
)] 𝑄𝑙𝑜𝑠𝑠 = ∑ 𝐵𝑖𝑗 (𝑉𝑖2 + 𝑉𝑗2 − 2𝑉𝑖 𝑉𝑗 𝑐𝑜𝑠𝛿𝑖𝑗 )
where: 𝑖=1
𝐵𝑖𝑗 susceptance of a line.
𝑃𝐺 , active power generation at the PV buses except
at the slack bus ; 𝑉𝐺 , voltage magnitude at PV
buses ; 𝑄𝐶 , shunt VAR compensation ; 𝑇 , tap 2.3. Constraints
settings of transformer; 𝑁𝐶, number of VAR The equality constraints represent the balance
between generated and load powers:
compensators ; 𝑁𝑇 , number of regulating 𝑛
(13)
transformers.
𝑃𝐺𝑖 − 𝑃𝐷𝑖 = |𝑉𝑖 | ∑|𝑉𝑗 |(𝐺𝑖𝑗 𝑐𝑜𝑠𝛿𝑖𝑗
𝑗=1
2.2. Objectives functions
Case 1: Generation fuel cost + 𝐵𝑖𝑗 𝑠𝑖𝑛𝛿𝑖𝑗 )
𝑛
The objective function is the quadratic equation of (14)
total generation fuel cost, can be expressed as: 𝑄𝐺𝑖 − 𝑄𝐷𝑖 = |𝑉𝑖 | ∑|𝑉𝑗 |(𝐺𝑖𝑗 𝑐𝑜𝑠𝛿𝑖𝑗
𝑁𝐺 𝑁𝐺 𝑗=1

𝐹1 = ∑ 𝐹𝑖 (𝑃𝐺𝑖 ) = ∑(𝑎𝑖 + 𝑏𝑖 𝑃𝐺𝑖 + 2

𝑐𝑖 𝑃𝐺𝑖 ) (6) + 𝐵𝑖𝑗 𝑠𝑖𝑛𝛿𝑖𝑗 )
𝑖=1 𝑖=1 The inequality constraints, also called security
𝑎𝑖 , 𝑏𝑖 , 𝑐𝑖 are the cost coefficients of 𝑖th generator. constraints, are directly associate with limits of
system:
𝑚𝑖𝑛 𝑚𝑎𝑥 (15)
Case 2: Voltage profile improvement 𝑃𝐺𝑖 ≤ 𝑃𝐺𝑖 ≤ 𝑃𝐺𝑖 𝑖 = 1,2, … , 𝑁𝐺
Objective is minimizing the voltage 𝑚𝑖𝑛 𝑚𝑎𝑥 (16)
𝑉𝐺𝑖 ≤ 𝑉𝐺𝑖 ≤ 𝑉𝐺𝑖 𝑖 = 1,2, … , 𝑁𝐺
deviations of load buses from a specified voltage. 𝑚𝑖𝑛
𝑄𝐺𝑖 ≤ 𝑄𝐺𝑖 ≤ 𝑄𝐺𝑖𝑚𝑎𝑥
𝑖 = 1,2, … , 𝑁𝐺 (17)
𝑁𝑃𝑄 𝑚𝑖𝑛 𝑚𝑎𝑥
𝑇𝑖 ≤ 𝑇𝑖 ≤ 𝑇𝑖 𝑖 = 1,2, … , 𝑁𝑇 (18)
𝐹3 = 𝑉𝐷 = ∑|(𝑉𝑖 − 1)| (7) 𝑚𝑖𝑛
𝑄𝐶𝑖 ≤ 𝑄𝐶𝑖 ≤ 𝑄𝐶𝑖 𝑚𝑎𝑥
𝑖 = 1,2, … , 𝑁𝐶 (19)
𝑖=1 𝑚𝑖𝑛 (20)
𝑆𝐿𝑖 ≤ 𝑆𝐿𝑖 𝑖 = 1,2, … , 𝑁𝑇𝐿
Case 3: Voltage stability enhancement 𝑉𝐿𝑖𝑚𝑖𝑛 ≤ 𝑉𝐿𝑖 ≤ 𝑉𝐿𝑖𝑚𝑎𝑥 𝑖 = 1,2, … , 𝑁𝑃𝑄 (21)
According to Kessel et al. [30], this objective
function can be formulated by minimizing voltage 2.4. DTBO
stability index 𝐿𝑚𝑎𝑥 : DTBO is human inspired metaheuristic
𝐹4 = 𝑚𝑖𝑛(𝐿𝑚𝑎𝑥 ) = 𝑚𝑖𝑛(𝑚𝑎𝑥(𝐿𝑛 )) algorithm developed by Mohammad Dehghani et al.
(8) (2022) in [31] where inspiration and main idea are
𝑛 = 1,2, … , 𝑁𝑃𝑄
𝑁𝑃𝑉 (9) explained as follow: “Driving training is an
𝑉𝑖 intelligent process in which a learner driver is
𝐿𝑛 = |1 − ∑ 𝐹𝑖𝑗 |
𝑉𝑗 trained and acquires driving skills. Learner driver
𝑖=1
Voltage stability index vary between 0 to 1 (no load can choose from several instructors when attending
case to voltage collapse case). driving school. The instructor then teaches the
learner driver the instructions and skills. The learner

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 850
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

driver tries to learn driving skills from the instructor In each iteration 𝑡, the best members are considered
and drive following the instructor. In addition, as driving instructors and the rest as learner drivers.
personal practice can improve the driver’s skills of The number of driving instructors 𝑁𝐷𝐼 decrements
the learner. These interactions and activities have for each iteration by using the following equation:
extraordinary potential for designing an optimizer”. 𝑁𝐷𝐼 = 0.1 ∙ 𝑁 ∙ (1 − 𝑡/𝑇 ) (26)
Population of DTBO is composed of 𝑇 is the maximum number of iterations.
driving learners and instructors where size 𝑁 is The new position for each member is calculated
selected manually. DTBO members are candidate using next equation:
solutions to the OPF problem modelled by the 𝑃1
𝑥𝑖,𝑗
population matrix:
𝑋1 𝑥𝑖,𝑗 + 𝑟 ∙ (𝐷𝐼𝑘𝑖,𝑗 − 𝐼 ∙ 𝑥𝑖,𝑗 ) , 𝐹𝐷𝐼𝑘 < 𝐹𝑖 (27
𝑖
={ )
⋮ 𝑥𝑖,𝑗 + 𝑟 ∙ (𝑥𝑖,𝑗 − 𝐷𝐼𝑘𝑖,𝑗 ) , 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒.
𝑋 = 𝑋𝑖
⋮ 𝐼 is a number randomly selected from the set [1,2],𝑟
[𝑋𝑁 ]𝑁×1 is a random number in the interval [0,1].
𝑥11 ⋯ 𝑥1𝑗 ⋯ 𝑥1𝑚 (22) This new position replaces the previous one if it
⋮ ⋱ ⋮ ⋱ ⋮ improves the value of the objective function.
= 𝑥𝑖1 ⋯ 𝑥𝑖𝑗 ⋯ 𝑥𝑖𝑚 𝑋 𝑃1 , 𝑠𝑖𝐹𝑖𝑃1 < 𝐹𝑖
⋮ ⋱ ⋮ 𝑋𝑖 = { 𝑖 (28)
⋮ ⋱ 𝑋𝑖 , 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
[𝑥𝑁1 ⋯ 𝑥𝑁𝑗 ⋯ 𝑥𝑁𝑚 ]𝑁×𝑚 Phase 2: Patterning of the instructor skills of the
𝑚 is the number of problem variables, 𝑋𝑖 is the 𝑖th student driver (exploration)
candidate solution, 𝑥𝑖𝑗 is the value of the 𝑗th The second phase is based on the learner driver
variable determined by the 𝑖th candidate solution imitating the instructor. This process moves
and is randomly initialized using following population to different positions in the search space,
equation: thus increasing the DTBO’s exploration power.
𝑃2
𝑥𝑖𝑗 = 𝑙𝑏𝑗 + 𝑟 ∙ (𝑢𝑏𝑗 − 𝑙𝑏𝑗 ), 𝑥𝑖,𝑗 = 𝑃 ∙ 𝑥𝑖,𝑗 + (1 − 𝑃) ∙ 𝐷𝐼𝑘𝑖,𝑗 (29)
(23) 𝑃2 𝑃2
𝑖 = 1,2, … , 𝑁, 𝑗 = 1,2, … , 𝑚 𝑋 , 𝑠𝑖 𝐹𝑖 < 𝐹𝑖
𝑋𝑖 = { 𝑖 (30)
𝑙𝑏𝑗 and 𝑢𝑏𝑗 are respectively the lower and upper 𝑋𝑖 , 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
bounds of the𝑗th variable;𝑟is a random number from 𝑃is the patterning index given by
[0,1]. 𝑃 = 0.001 + 0.9(1 − 𝑡/𝑇) (31)
Each candidate is placed in the objective function, Phase 3: Personal practice (exploitation)
and the member that has the best value is known as This third phase is based on the personal practice of
the best member of the population (𝑋𝑏𝑒𝑠𝑡 ). each learner driver to improve and enhance driving
𝐹1 𝐹(𝑋1 ) skills.
⋮ ⋮ 𝑃3
𝑥𝑖,𝑗 = 𝑥𝑖,𝑗 + (1 − 2𝑟) ∙ 𝑅 ∙ (1 − 𝑡/𝑇) ∙ 𝑥𝑖,𝑗 (32)
𝐹 = 𝐹𝑖 = 𝐹(𝑋𝑖 ) (24) 𝑃3
𝑋 , 𝑃3
𝑠𝑖 𝐹𝑖 < 𝐹𝑖 (33)
⋮ ⋮ 𝑋𝑖 = { 𝑖
𝑋𝑖 , 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
[𝐹𝑁 ]𝑁×1 [𝐹(𝑋𝑁 )]𝑁×1
𝑟 is a random real number of the interval [0,1], 𝑅 is
DTBO is modelled in three different phases: the constant set to 0.05.

2.5. MDTBO
MDTBO introduce two modifications of original
Phase 1: Training by the driving instructor DTBO method. Firstly, size of population is fixed at
(exploration) the same value of problem’s dimension.
The first phase consists on the choice of the driving 𝑁=𝑚 (34)
instructor by the learner driver and then the training Secondly, the number of driving instructors in first
of the driving by the selected instructor to the phase is updated to the following equation:
learner driver. It can be expressed as: 𝑁𝐷𝐼 = 1 + 0.2 ∙ 𝑁 ∙ (1 − 𝑡/𝑇 ) (35)
𝐷𝐼 Member of driving instructors decrement slowly for
𝐷𝐼11 ⋯ 𝐷𝐼1𝑗 ⋯ 𝐷𝐼1𝑚 each iteration, but it not gets zero value until the end
⋮ ⋱ ⋮ ⋱ ⋮ (25 of iteration. Phase of exploration is improved. Fig. 2
= 𝐷𝐼𝑖1 ⋯ 𝐷𝐼𝑖𝑗 ⋯ 𝐷𝐼𝑖𝑚 ) shows variation of 𝑁𝐷𝐼 for 𝑁 = 24 and 𝑇 = 500
⋮ ⋱ ⋮ ⋱ ⋮ width DTBO and MDTBO.
[𝐷𝐼𝑁𝐷𝐼1 ⋯ 𝐷𝐼𝑁𝐷𝐼𝑗 ⋯ 𝐷𝐼𝑁𝐷𝐼𝑚 ]𝑁 ×𝑚
𝐷𝐼

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 851
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

Fig.2: Variations of the number of driving instructors with MDTBO and DTBO

Flowchart of MDTBO is presented in the following figure.

Fig.3: Flowchart of MDTBO

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 852
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

III. RESULTS simulation, with Intel Core i5-3230M CPU @

For solving OPF problem, MDTBO 2.60GHz processor and 8GO RAM.
algorithm has been tested on the standard IEEE 30- The variation of fuel cost is shown in Fig.4
bus system using Matlab software for coding and for case 1 (minimization of fuel cost), with
MDTBO, TLBO and PSO algorithms.

Fig.4: Fuel cost variation for case 1 with MDTBO, TLBO and PSO

For case 2, voltage profile improvement,

variations of fuel cost and voltage deviations (VD)
with MDTBO approach is presented in Fig. 5.

Fig.5: Variations of fuel cost and voltage deviations for case 2 with MDTBO

Result of voltage stability enhancement for fuel cost and voltage stability index 𝐿𝑚𝑎𝑥
(case 3), is shown in Fig. 6, where MDTBO is used variations.
DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 853
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

Fig.6: Variations of fuel cost and voltage stability index for case 3 with MDTBO

Fig. 7 and Fig. 8 represent, respectively, case 5. Results of MDTBO is compared to TLBO
the variations of active power transmission losses and PSO.
𝑃𝑙𝑜𝑠𝑠 for case 4, and reactive power losses 𝑄𝑙𝑜𝑠𝑠 for

Fig.7: Variations of active power transmission losses for case 4 with MDTBO, TLBO and PSO

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 854
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

Fig.8: Variations of reactive power losses for case 5 with MDTBO, TLBO and PSO

With MDTBO approach, voltage magnitude of 30-bus system for case 1 is compared to case 2 in Fig.9.

Fig.9: System voltage profiles for case 1 and case 2 with MDTBO

Active power transmission losses for case 4 Reactive power losses for case 5 is also compared to
is compared to case 1 and illustrate in Fig. 10. case 1 and illustrate in Fig. 11.

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 855
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

Fig.10: Active power transmission losses for case 1 and case with MDTBO

Fig.11: Reactive power losses for case 1 and case 5 with MDTBO

Optimal settings of control variables and there are some voltage violations at bus 19 through
elapsed times for case 0 (initial case), case 1, case 2, bus 30, but there are no longer violations in the
case 3, case 4 and case 5, with MDTBO, TLBO and results obtained using MDTBO (Fig. 9). In Fig. 4, it
PSO, are given in Table 3 and Table 4. appears that the convergence of the proposed
method is competitive with TLBO and PSO.
IV. DISCUSSION In case 2, width MDTBO, the total
Initially, the total fuel cost was $901.9501. generation fuel cost and voltage deviations are
The total cost obtained by MDTBO in case 1 is $804.1092 and 0.1322 p.u. compared to $799.5753
$799.5753, while with PSO it is $799.5823. Fuel and 1.4079 p.u. for case 1. So the cost has been
cost is greatly reduced (11.35% reduction) and increased by 0.57%, and the voltage profile has been
MDTBO is more accurate than PSO. With TLBO, improved by 90.61% with MDTBO (Fig. 5).
fuel cost in case 1 has a best value $799.0680, but Voltage profile improvement is interesting with
elapsed time (305.0494s) is not interesting TLBO but execution time of one iteration is not.
compared to MDTBO (177.6250s). In initial case,

Table 3: Optimal settings of control variables for case 1 and case 2, with PSO, TLBO and MDTBO
Case 1 Case 2
Variable Mi Ma Case 1 Case 1 Case 2 Case 2
Case 0 MDTB MDTB
s n x PSO TLBO PSO TLBO
O O
𝑃1 50 200 99.2225 177.301 177.057 176.934 181.368 176.361 174.367
𝑃2 20 80 80 48.7654 48.6972 48.6810 45.7325 48.8767 48.5661
𝑃5 15 50 50 21.3155 21.3043 21.2255 22.1605 21.6590 21.1154
𝑃8 10 35 20 20.8126 21.0814 20.9396 19.2589 22.1195 21.2827
𝑃11 10 30 20 11.8358 11.8842 11.7209 12.5495 12.1699 13.5531
𝑃13 12 40 20 12.1574 12.000 12.6520 12.5255 12.0001 14.2735
𝑉1 0.95 1.1 1.0500 1.1000 1.1000 1.1000 1.0370 1.0401 1.0397
𝑉2 0.95 1.1 1.0400 1.0873 1.0879 1.0876 1.0199 1.0237 1.0237
𝑉5 0.95 1.1 1.0100 1.0613 1.0617 1.0608 1.0190 1.0140 1.0143

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 856
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

𝑉8 0.95 1.1 1.0100 1.0695 1.0694 1.0689 1.0016 1.0046 1.0083

𝑉11 0.95 1.1 1.0500 1.0999 1.1000 1.1000 1.0528 1.0606 1.0141
𝑉13 0.95 1.1 1.0500 1.0999 1.1000 1.1000 0.9997 0.9876 0.9984
𝑇11 0.9 1.1 1.0780 0.9902 1.0447 0.9497 1.0051 1.0833 0.9793
𝑇12 0.9 1.1 1.0690 1.0436 0.9000 1.0125 0.9311 0.9012 0.9272
𝑇15 0.9 1.1 1.0320 1.0999 0.9863 1.0177 0.9653 0.9379 0.9506
𝑇36 0.9 1.1 1.0680 1.0123 0.9657 0.9717 0.9736 0.9707 0.9582
𝑄10 0.0 5.0 0 0.0018 5.000 2.2939 1.7482 4.8524 2.4286
𝑄12 0.0 5.0 0 4.8882 5.000 1.5445 4.8098 0 2.7292
𝑄15 0.0 5.0 0 1.4461 5.000 2.1016 4.2538 4.9999 3.3897
𝑄17 0.0 5.0 0 4.9987 5.000 2.4704 0.0887 0 2.5774
𝑄20 0.0 5.0 0 1.8570 5.000 0.8641 4.8249 5 2.3691
𝑄21 0.0 5.0 0 0.0004 5.000 3.6410 2.6886 4.9998 4.5337
𝑄23 0.0 5.0 0 4.9983 3.8490 1.7289 4.3941 5 5
𝑄24 0.0 5.0 0 3.1354 5.000 0.77854 3.9468 5 1.3463
𝑄29 0.0 5.0 0 4.9907 2.7434 1.0707 4.7485 2.6146 1.7527
Cost
901.950 799.582 799.068 799.575 804.346 803.460 804.109
($/h)
𝑃𝑙𝑜𝑠𝑠
5.8225 8.7880 8.6245 8.7532 10.2958 9.8872 9.8513
(MW)
𝑄𝑙𝑜𝑠𝑠
-4.6063 3.2444 4.1827 5.0820 14.1149 12.3560 11.4872
(MVAR)
𝑉𝐷 1.1496 1.0757 1.8583 1.4079 0.11969 0.0945 0.1322

Lmax 0.1723 0.1270 0.1164 0.1268 0.13649 0.1369 0.1390

Elapsed
107.916 305.049 177.625 110.503 293.685 177.895
time

We can notice from Fig. 9 that the voltage problem in electric power systems that is MDTBO.
profile has been greatly improved with MDTBO. Original DTBO is a human-inspired metaheuristic
In case 3, voltage stability index is algorithm, based on imitating human actions when
significantly reduced compared to case 1 and case 2. learning to drive. It has three update phases: training
Fig. 6 shows that voltage stability is greatly by the driving instructor, patterning of the instructor
improved while optimizing generation of fuel cost. skills of the student driver, and personal practice.
Active power transmission losses are The MDTBO is a modified version of original
greatly reduced in case 4 with MDTBO (Total of DTBO, where population is set to the dimension of
8.7532MWfor case 1 to 2.9162MW for case 4), and the problem and number of driving instructor in first
it is illustrated in Fig.10. Fig.7 shows that phase is updating to avoid convergence to none
convergence of the proposed method is interesting instructor. Five objective functions have been
compared to TLBO and PSO. considered to minimize the fuel cost, to improve the
Fig.11 demonstrates that Reactive power losses are voltage profile, to enhance voltage stability, and to
also reduced in case 5 compared to case 1. Finally, minimize active power transmission losses and
with Fig. 8 we can say that MDTBO is more reactive power losses. The proposed method has
interesting than PSO, and very competitive with been tested on IEEE 30-bus system. Compared to
TLBO. TLBO and PSO, the results width MDTBO confirm
the potential of proposed approach and show his
V. CONCLUSION robustness and effectiveness to solve OPF problem.
In this paper, we have introduced and
presented a novel technique to solve the OPF

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 857
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

Table 4: Optimal settings of control variables for case 3, case 4 and case 5, with PSO, TLBO and MDTBO
Case 3 Case 4 Case 5
Case 3 Case 3 Case 4 Case 4 Case 5 Case 5
Variables MDTB MDTB MDTB
PSO TLBO PSO TLBO PSO TLBO
O O O
𝑃1 182.767 174.660 155.972 51.2654 51.2493 51.3149 51.8113 51.4494 52.1202
𝑃2 50.9850 49.8639 47.6195 80 80 80 80 80 80
𝑃5 21.6500 21.5708 25.3178 50 50 50 50 50 50
𝑃8 12.5835 20.9978 26.5968 35 35 35 35 35 35
𝑃11 11.3407 12.7767 17.0993 30 30 30 30 30 29.3955
𝑃13 13.7921 12.1572 18.1837 40 40 40 40 40 39.9990
𝑉1 1.0934 1.1 1.1 1.1000 1.1 1.1 1.1 1.1 1.1
𝑉2 1.0696 1.0882 1.0896 1.1000 1.0976 1.0987 1.1 1.1 1.1
𝑉5 1.0678 1.0758 1.0876 1.0827 1.0799 1.0813 1.1 1.0928 1.0928
𝑉8 1.0582 1.0870 1.0878 1.0898 1.0869 1.0887 1.1 1.1 1.1
𝑉11 1.0678 1.0995 1.0903 1.1000 1.1 1.0947 1.0454 1.0371 1.0378
𝑉13 1.0762 1.0999 1.0996 1.1000 1.1 1.0919 1.1 1.0644 1.0651
𝑇11 0.9267 0.9826 0.9433 1.1000 1.0545 0.9955 1.0452 1.0788 1.0754
𝑇12 0.9984 0.9211 1.0084 0.9000 0.9 1.0097 1.1 1.0245 1.0187
𝑇15 0.9865 0.9789 1.0096 0.9920 0.9842 1.0188 1.0091 1.0264 1.0228
𝑇36 0.9352 0.9749 0.9553 0.9804 0.9726 0.9957 1.0456 1.0490 1.0530
𝑄10 4.0461 0.0033 4.9988 5 5 3.3467 0 5 4.6580
𝑄12 2.2936 4.9985 1.4813 4.9997 5 3.7073 5 5 4.8339
𝑄15 4.2379 4.9825 1.9031 4.9875 4.9999 4.8351 5 5 4.9675
𝑄17 4.4928 4.9669 3.3039 5 5 3.4718 5 5 2.5841
𝑄20 3.8630 5 1.1050 5 4.8436 4.4972 0 5 4.6726
Q21 4.4736 4.9984 4.9856 5 5 4.0682 0 5 4.9976
Q23 1.7748 4.9840 3.4176 3.6861 3.6553 4.5559 0 5 4.8840
Q24 4.0680 4.9994 0.2098 5 5 2.3341 5 5 1.8897
Q29 4.8952 4.9993 3.0800 2.3548 2.5142 4.1913 5 .1496 4.5322

Cost ($/h) 802.042 799.64 805.332 967.103 999.971 967.221 968.406 967.542 966.428

Ploss (MW) 9.7183 8.6280 7.3901 2.8669 2.8506 2.9162 3.4133 3.0513 3.1164

Qloss
12.2530 4.6505 -0.8028 -20.131 -18.567 -20.281 -22.924 -24.213 -24.098
(MVAR)

VD 1.6006 2.1242 1.8505 1.9314 2.0502 1.6784 1.0637 1.0679 1.0211

Lmax 0.1161 0.1139 0.1172 0.1163 0.1150 0.1196 0.1291 0.1273 0.1285

Elapsed
113.255 309.208 177.121 114.205 304.039 183.019 112.637 305.913 173.800
time

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 858
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

ACKNOWLEDGMENT flow problem”, Int. J. Electr. Power

The corresponding author would like to Energy Syst. 53, 219-230.
thank Prof. Harlin Andriatsihoarana, Thesis [9]. Kahourzade, S.; Mahmoudi, A.; Mokhlis,
Director, and Edmond Randriamora, Co-Director, H.B, 2015, “A comparative study of
for their supervision and advice. As this publication multi-objective optimal power flow based
is the result of research works carried out at the on particle swarm, evolutionary
Doctoral School in Sciences and Techniques of programming, and genetic algorithm”,
Engineering and Innovation (ED-STII) of the Elect. Eng. 97, 1-12.
University of Antananarivo, Madagascar, so thanks [10]. Le Anh, T. N. ; Vo, D. N.; Ongsakul, W.,
to the Director of the ED-STII for having 2015, “Cuckoo Optimization Algorithm
welcomed him to this Doctoral School, and thanks for Optimal Power Flow”, Proceedings of
to all member of Electrical Engineering the 18th Asia Pacific Symposium on
Department for their encouragement. Intelligent and Evolutionary Systems,
Volume 1. 2025. 479-493.
REFERENCES [11]. El-Fergany, A.A.; Hasanien, H.M, 2015,
[1]. Carpentier, J., 1962, “Contribution à “Signle and multi-objective optimal power
l’étude du dispatching économique”, Bull. flow using grey wolf optimizer and
Soc. Franc. Elec. 3(1962) 431-447. differential evolution algorithms”, Electr.
[2]. Ebeed, M.; Kamel, S.; Jurado, F., 2018, Power Compon. Syst, 43, 1548-1559.
“Optimal Power Flow Using Recent [12]. Bouchekara, H.; Abido, M.; Boucherma,
Optimization Techniques”, Classical and M., 2014, “Optimal power flow using
Recent Aspects of Power System teaching-learning based optimization
Optimization, Elsevier Inc. (2018) 157- technique”, Electr. Power Syst. Res. 114,
183. 49-59.
[3]. Momoh, J.A.; Adapa, R.; El-HawaryM., [13]. Shabanpour-Haghighi, A.; Seifi, A.R.;
1999, “A review of selected optimal Niknam, T., 2014, “A modified teaching-
power flow literature to 1993. I. Nonlinear learning based optimization for multi-
and quadratic programming approaches”, bojective optimal power flow problem”,
IEEE Trans. Power Syst. 14(1999) 96- Energy Convers. Manag. 77, 597-607.
104. [14]. Ghasemi, M.; Ghavidel, S.; Gitizadeh, M.;
[4]. Momoh J.A; Adapa, R.; El-Hawary, M., Akbari, E., 2015, “An improved teaching-
1999, “A review of selected optimal learning based optimization algorithm
power flow literature to 1993. II. Newton, using Levy mutation strategy for non-
linear programming and interior point smooth optimal power flow”, Int. J. Electr.
methods”, IEEE Trans. Power Syst. Power Energy Syst. 65, 375-384.
14(1999) 105-111. [15]. Duman, S., 2016, “Symbiotic organisms
[5]. Emmanuel, A.Nyiekaa; search algorithm for optimal power flow
IkennaOnyegbadue, 2021, “Optimal problem based on valve-point effect and
Power Flow Mehtods – A Survey”, prohibited zones”, Neural Comput. Appl.
International Journal of Advances in 1-15.
Engineering and Management (IJAEM), [16]. Warid, W., 2020, “Optimal power flow
Volume 3, Issue 2 Feb 2021, pp: 20-32. using the AMTPG-Jaya algorithm”,
[6]. Risi, B.G; Riganti-Fulginei, F.; Laudani, Applied Soft Computing Journal, 106252.
A., 2022, “Modern Techniques for the [17]. Warid, W., 2022, “A Novel Chaotic Rao-2
Optimal Power Flow Problem, State of the Algorithm for Optimal Power Flow
Art”, Energies 2022, 15, 6387. Solution”, Journal of Electrical and
[7]. Attia, A.-F.; Al-Turki, Y.A.; Abusorrah, Computer Engineering, vol. 2022, Article
A.M., 2012, “Optimal power flow using ID 7694026, 19 pages.
adapted genetic algorithm width adjusting [18]. Mohamed, A.-A.A.; Mohamed, Y.S.; El-
population size”, Electr. Power Compon. Gaafary, A.A.; Hemeida, A.M., 2017,
Syst. 40, 1285-1299. “Optimal power flow using moth swarm
[8]. Adaryani, M.R.; Karami, A., 2013, algorithm”, Electr. Power Syst. Res. 142,
“Artificial bee colony algorithm for 190-206.
solving multi-objective optimal power [19]. Li, Z.; Cao, Y.; Dai, L.V.; Yang, X. ;
Nguyen, T.T., 2019, “Optimal power flow

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 859
 International Journal of Advances in Engineering and Management (IJAEM)
Volume 7, Issue 02 Feb. 2025, pp: 846-860 www.ijaem.net ISSN: 2395-5252

for transmission power networks using a power system”, IEEE Trans. Power Deliv.
Novel Metaheuristic Algorithm”, 1, 346-354.
Energies, 12(22), 4310. [31]. Mohammad Dehghani; Eva Trojovská;
[20]. Shaheen, A.; Ginidi, A.; El-Sehiemy, R.; Pavel Trojovský, 2022, “A new human-
Elsayed, A.; Elattar, E.; Dorrah, H.T, based metaheuristic algorithm for solving
2022, “Developed Gorilla troops optimization problems on the base of
technique for optimal power flow problem simulation of driving training process”,
in electrical power systems”, Scientific reports 2022.
Mathematics, vol. 10, p.1636.
[21]. Nguyen, T.T; Duong, MQ.; et al., 2020,
“Optimal operation of transmission power
networks by using improved stochastic
fractal search algorithm”, Neural
Comput&Applic. 32, 9129-9164.
[22]. Yuan, X.; Zhang, B; Wang, P; Liang, J;
Yuan, Y., Huang, Y., et al., 2017, “Multi-
objective optimal power flow based on
improved strength Pareto evolutionary
algorithm”, Energy 122, 70-82.
[23]. Ramesh, V.C; Li, X., 1997, “A fuzzy
multiobjective approach to contingency
constrained OPF”, IEEE Trans. Power
Syst. 12, 1348-1354.
[24]. Venkatesh, B., 2003, “Online ANN
memory model-based method for unified
OPF and voltage stability margin
maximization, Electr. Power Compon.
Syst. 31, 453-465.
[25]. Yuhao, Z.; Bei, Z.; Chunlei, X.; Tu, L.;
Ruisheng, D.; Di Shi; Zhiwei, W.; Wei-
Jen, L., 2020, “A Data-driven Method for
Fast AC Optimal Power Flow Solutions
via Deep Reinforcement Learning”,
Journal of modern power systems and
clean energy, vol. 8, NO. 6.
[26]. Singh, M. K.; Kekatos, V.; Giannakis,
G.B, 2022, “Learning to solve the AC-
OPF using sensitivity-informed deep
neural networks”, IEEE Trans. Power
Syst., 37(4):2833-2846.
[27]. Xiang Pan; Minghua Chen; Tianyu Zhao;
Steven, H., 2023, “Low, DeepOPF: a
feasibility-optimized deep neural network
approach for AC optimal power flow
problems”, IEEE Systems Journal, 17(1),
673-683.
[28]. Abou El Ela, A.A.; Abido, M.A.; 2010,
“Optimal power flow using differential
evolution algorithm”, Electr. Power Syst.
Res. 80(7), 878-885.
[29]. Alsac, O.; Stot, B; 1974, “Optimal load
flow with steady-state security”, IEEE
Trans. Power App. Syst. PAS-93 (3), 745-
751.
[30]. Kessel, P.; Glavitsch, H., 1986,
“Estimating the voltage stability of a

DOI: 10.35629/5252-0702846860 |Impact Factorvalue 6.18| ISO 9001: 2008 Certified Journal Page 860